On Tuesday, November 23, 2010 01:21:45 Laurent Pinchart wrote:
> Hi Hans,
>
> On Tuesday 23 November 2010 01:14:28 Hans Verkuil wrote:
> > On Monday, November 22, 2010 12:20:25 Hans Verkuil wrote:
> > > On Monday, November 22, 2010 11:21:27 Laurent Pinchart wrote:
> > > > Removing the mutex_lock/unlock around clear_bit allowed
> > > > device_unregister() to race with v4l2_open(). The device can be
> > > > unregistered between the video_devdata() and video_get() calls.
> > > >
> > > > Revert the patch to fix the problem.
> > > >
> > > > This reverts commit dd0daf2a6fb6bec436a3ef68bd585ea09a2a54b7.
> > > >
> > > > Signed-off-by: Laurent Pinchart <laurent.pinch...@ideasonboard.com>
> > >
> > > Acked-by: Hans Verkuil <hverk...@xs4all.nl>
> >
> > Change of mind, I nack this.
> >
> > I realized that the mutex should be around device_unregister, not around
> > the clear_bit. The mutex protects access to vdev->dev and not vdev->flags
> > since that's atomic anyway.
>
> device_unregister() will end up calling v4l2_device_release(), which also
> locks videodev_lock. That will result in a deadlock.
True. (Note to myself: don't post past midnight...)
But a mutex around clear_bit doesn't help either. Unless v4l2_open will test
for this bit just before the video_get(). E.g. by replacing 'if (vdev == NULL)'
with 'if (vdev == NULL || !video_is_registered(vdev))'.
I think that will do the trick.
Regards,
Hans
--
Hans Verkuil - video4linux developer - sponsored by Cisco
--
To unsubscribe from this list: send the line "unsubscribe linux-media" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
